To be able to continually increase the gate count of semiconductor devices on fixed die size, integrated circuit (IC) designs have involved shrinking feature sizes. For the next decade, the outlook is strong for photolithography to continue to be the process by which IC are manufactured. When processing the features in today's deep sub-micron processes, the wavelength of light used in the photolithography process is less than that of the feature size. A result of the use of photolithography under these “tight” conditions is that the resulting design, notwithstanding the use of phase shift masking, does not precisely match the desired design.
A method of automatically correcting the resulting differences involves making subtle modifications to the mask or reticle used in the photolithography process (hereinafter collectively referred to as mask). These modifications are termed optical proximity corrections or optical and process corrections. Whether the term is referring to optical distortions alone or for process distortions in addition to optical distortions determines which term is the proper term to use. Regardless of the reason for these corrections, the discussions herein will generically refer to either or both of these types of corrections as OPC.
There are two basic types of OPC, rule-based and model-based. Rule based OPC applies corrections to the mask based on a predetermined set of rules. Thus, if an analysis of the mask determines that the mask meets a predetermined set of conditions, a process applies the appropriate correction to the mask for the conditions met. The corrections resulting from the rule-based approaches are typically less accurate, when compared to model based correction. However, rule-based corrections are more computationally efficient, and less costly. In contrast, a model-based OPC technique uses process simulation to determine corrections to the masks. The model-based OPC corrections, generated in accordance with the results of these simulations, generally provide for greater accuracy than the corrections provided by rule-based OPC. However, model-based OPC is computationally intensive and therefore time consuming as well as costly.